Family Satisfaction Survey

• Objective: To establish the validity of a total score for the Critical Care Family Satisfaction Survey (CCFSS). • Design: Instrument validation study. • Setting and participants: 2494 family members of patients admitted to 10 critical care units over a 3-year period. • Methods: A total score was computed by summing the 5 subscales of the survey, providing a score in the range of 5 to 25. Total score validation techniques included comparison of a model developed from the recently collected survey data with the model developed for initial validation of the survey (n = 145). Goodness-of-fit analyses included calculation and comparison of internal consistency estimates (Cronbach’s alpha) from both models using Fisher’s r-to-z transformations as well as construct validation using r-to-z transformation and comparisons of standardized regression weights from both models. A scree plot was generated to demonstrate the similarity between the model’s eigenstructures, and a chi-square goodness-of-fit analysis was performed to compare the percentage of variance explained by each model. • Results: Internal consistency estimates from the 2 models were not significantly different (z score P values > 0.10). Statistical differences between standardized regression weights were observed, and these were caused by the large sample size in this study. However, the patterns of differences between these weights across the 4and 5-construct model were nearly identical. Goodness-of-fit analysis of the percentage of variance explained by each model demonstrated agreement of the underlying eigenstructure of the models (P = 0.80). • Conclusion: Comparison of the models showed significant agreement and no meaningful differences, demonstrating the validity of a total score for the CCFSS. Evidence of a single latent construct being estimated from both models provides additional support for the computation of a total score. Patient satisfaction is recognized as an important outcome of hospital admission [1,2], and quality improvement efforts must consider patient satisfaction in order to be fully comprehensive [3]. Studies of patient satisfaction in the critical care setting have shown that overall satisfaction with care extends to the family [4,5]. This is true for 2 reasons. Decision making for the patient is the responsibility of family members [5], and patients’ varying levels of consciousness and the severity of their condition may alter or confuse their perception of care [1]. Hospitals also are recognizing that the needs of patients and their families represent constructs separate from satisfaction with care [5]. The Critical Care Family Satisfaction Survey (CCFSS) was developed as a tool to measure family and patient satisfaction within the critical care setting [1,6] and has been selected by the Joint Commission on Accreditation of Healthcare Organizations as a potential core measure. The CCFSS contains 20 items stratified into 5 constructs (Assurance, Proximity, Information, Comfort, and Support) with a score calculated for each construct. Since the instrument was developed, we have concluded that a total overall score would be appropriate and meaningful for several reasons. A single score could be used to quickly represent and monitor family satisfaction and would be more useful in many applications than a close examination of the 5 validated subscales. In addition, a total score would facilitate research and repeated-measures monitoring of outcomes of care (where satisfaction is concerned) over time. Finally, researchers could incorporate a total score into more complex statistical models and perform other data analyses more easily. The calculation of a total score was not part of the original design or statistical analysis for the instrument, although the total score concept was validated in the initial validation study [1]. In the original study, principal components analysis with generation of a scree plot, an exploratory factor analysis (EFA) technique, was performed to determine how From the Department of Health Studies and the Department of Medicine, Lehigh Valley Hospital, Allentown, PA, and the Department of Health Evaluation Sciences, Pennsylvania State University School of Medicine, Hershey, PA. many factors (ie, constructs) should remain in the model. This analysis yields eigenvalues representing the percentage of variance explained by each factor. Values greater than 1.0 are interpreted to indicate an independent factor [7]. The validation study found that a single factor had an eigenvalue of 3.712 and thus explained approximately 75% of the model variation. All other eigenvalues were less than 1.0. These results clearly indicate that the 5 subscales were estimating a single latent variable, presumably “satisfaction.” Using this logic, we developed this study to determine if a single total score using the sum of the 5 subscales could be validated. This article reports on the statistical calculations and procedures that were used to validate a total score for the CCFSS.